[Sci-Tech Institute NOW] GIST predicts immunotherapy response by analyzing individual cancer cells

A staff led by Professor Park Jihwan at GIST has developed a know-how to exactly predict the response to immuno-oncology remedy by analyzing the variations in immune response of individual cancer cells composing a tumor on the mobile degree. Image of colorectal cancer. Provided by Getty Images Bank

■ The Gwangju Institute of Science and Technology (GIST) introduced on the sixth {that a} staff led by Professor Park Jihwan of the Department of Life Sciences has developed ‘scMnT,’ an evaluation know-how that exactly predicts the response to immuno-oncology remedy on the single-cell degree. Conventional evaluation strategies solely decided Microsatellite Instability (MSI) as optimistic or adverse based mostly on the typical worth of all the tumor, making it obscure the detailed remedy response for every affected person. The analysis staff developed a technique to quantify MSI as a steady indicator with various depth and analyze it on the mobile degree. When utilized to colorectal cancer affected person information, the staff confirmed that in areas with excessive MSI depth, immune cells had been concentrated and attacked cancer cells, whereas in areas with low depth, the immune response was blunted. This is predicted to contribute to realizing precision medication tailor-made to every affected person’s tumor traits and enhancing the success price of immuno-oncology remedy.

■ POSTECH introduced on the sixth {that a} staff led by Professor Kim Yong-jun of the Department of Electrical and Electronic Engineering and the Graduate School of Convergence Science and Technology, in collaboration with Seoul National University, Sungkyunkwan University, and the University of Ulsan, has developed ‘EfficientMPT’ know-how. This know-how drastically reduces the excessive computational load and reminiscence necessities of present AI-based communication decoders. By designing the ‘consideration’ mechanism, a core computation course of in Transformer AI fashions, with light-weight vector operations as a substitute of advanced matrix calculations, it reduces reminiscence utilization by as much as 91% and computational load by as much as 57% in comparison with present AI decoders, whereas sustaining error correction efficiency. It additionally features as a basis mannequin that, as soon as educated, could be utilized to numerous sorts and lengths of information. It is predicted to be utilized in next-generation communication programs like 6G and AI-RAN, in addition to in SSD storage gadgets.

■ KAIST introduced on the sixth that its Kim Jaechul Graduate School of AI will maintain the ‘KAIST AI Tech Fair 2026’ at COEX in Seoul on the seventh. Co-hosted by the Seongnam Industry Promotion Agency and the Seoul Metropolitan Government, the occasion will introduce the newest analysis achievements to trade and most people, together with robotic basis fashions, customized AI, multimodal AI, reliable AI, and bodily AI. Anyone, together with company officers, researchers, and college students, can attend, and consultations for know-how switch and joint analysis can be obtainable after the occasion. For inquiries about know-how switch and joint analysis, contact the KAIST Seongnam Research Center (Director: Professor Choi Jaesik, 031-8022-7530, [email protected]).